Silicon wafers or sheets may be used in, for example, the integrated circuit or solar cell industry. Demand for solar cells continues to increase as the demand for renewable energy sources increases. As these demands increase, one goal of the solar cell industry is to lower the cost/power ratio. There are two types of solar cells: silicon and thin film. The majority of solar cells are made from silicon wafers, such as single crystal silicon wafers. Currently, a major cost of a crystalline silicon solar cell is the wafer on which the solar cell is made. The efficiency of the solar cell, or the amount of power produced under standard illumination, is limited, in part, by the quality of this wafer. Any reduction in the cost of manufacturing a wafer without decreasing quality will lower the cost/power ratio and enable the wider availability of this clean energy technology.
The highest efficiency silicon solar cells may have an efficiency of greater than 20%. These are made using electronics-grade monocrystalline silicon wafers. Such wafers may be made by sawing thin slices from a monocrystalline silicon cylindrical boule grown using the Czochralski method. These slices may be less than 200 μm thick. To maintain single crystal growth, the boule must be grown slowly, such as less than 10 μm/s, from a crucible containing a melt. The subsequent sawing process leads to approximately 200 μm of kerf loss, or loss due to the width of a saw blade, per wafer. The cylindrical boule or wafer also may need to be squared off to make a square solar cell. Both the squaring and kerf losses lead to material waste and increased material costs. As solar cells become thinner, the percent of silicon waste per cut increases. Limits to ingot slicing technology, however, may hinder the ability to obtain thinner solar cells.
Other solar cells are made using wafers sawed from polycrystalline silicon ingots. Polycrystalline silicon ingots may be grown faster than monocrystalline silicon. However, the quality of the resulting wafers is lower because there are more defects and grain boundaries and this lower quality results in lower efficiency solar cells. The sawing process for a polycrystalline silicon ingot is as inefficient as a monocrystalline silicon ingot or boule.
Another solution that may reduce silicon waste is cleaving a wafer from a silicon ingot after ion implantation. For example, hydrogen, helium, or other noble gas ions are implanted beneath the surface of the silicon ingot to form an implanted region. This is followed by a thermal, physical, or chemical treatment to cleave the wafer from the ingot along this implanted region. While cleaving through ion implantation can produce wafers without kerf losses, it has yet to be proven that this method can be employed to produce silicon wafers economically.
Yet another solution is to pull a thin ribbon of silicon vertically from a melt and then allow the pulled silicon to cool and solidify into a sheet. The pull rate of this method may be limited to less than approximately 18 mm/minute. The removed latent heat during cooling and solidifying of the silicon must be removed along the vertical ribbon. This results in a large temperature gradient along the ribbon. This temperature gradient stresses the crystalline silicon ribbon and may result in poor quality multi-grain silicon. The width and thickness of the ribbon also may be limited due to this temperature gradient. For example, the width may be limited to less than 80 mm and the thickness may be limited to 180 μm.
Horizontal ribbons of silicon that are physically pulled from a melt also have been tested. In one method, a seed attached to a rod is inserted into the melt and the rod and resulting sheet are pulled at a low angle over the edge of the crucible. The angle and surface tension are balanced to prevent the melt from spilling over the crucible. It is difficult, however, to initiate and control such a pulling process. Access must be given to the crucible and melt to insert the seed, which may result in heat loss. Additional heat may be added to the crucible to compensate for this heat loss. This may cause vertical temperature gradients in the melt that may cause non-laminar fluid flow. Also, a possibly difficult angle of inclination adjustment to balance gravity and surface tension of the meniscus formed at the crucible edge must be performed. Furthermore, since heat is being removed at the separation point of the sheet and melt, there is a sudden change between heat being removed as latent heat and heat being removed as sensible heat. This may cause a large temperature gradient along the ribbon at this separation point and may cause dislocations in the crystal. Dislocations and warping may occur due to these temperature gradients along the sheet.
Production of thin sheets separated horizontally from a melt, such as by using a spillway, has not been performed. Producing sheets horizontally from a melt by separation may be less expensive than silicon sliced from an ingot and may eliminate kerf loss or loss due to squaring. Sheets produced horizontally from a melt by separation also may be less expensive than silicon cleaved from an ingot using hydrogen ions or other pulled silicon ribbon methods. Furthermore, separating a sheet horizontally from a melt may improve the crystal quality of the sheet compared to pulled ribbons. A crystal growth method such as this that can reduce material costs would be a major enabling step to reduce the cost of silicon solar cells. Yet impurities or solutes may build up in the melt. Accordingly, there is a need in the art for an improved apparatus and method for purifying a melt.